Police struggle to identify the riskiest domestic abuse perpetrators – here’s how they can do better

The government cannot achieve its target to halve violence against women and girls if it doesn’t address the most serious perpetrators – and it isn’t anywhere near knowing how to identify them. Our new research shows where they are going wrong, and how they can do better.

The most recent statistics show that violence against women and girls affects one in 12 women in England and Wales. A quarter of domestic abuse incidents reported to police involve known, repeat perpetrators. But despite being told by government to identify and control the most serious perpetrators, police do not currently have systems good enough to do that.

Currently, police forces use an algorithm to determine which offenders pose the greatest risk to women and girls. This is known as the RFGV algorithm – perpetrators are propelled up or down a list based on the recency, frequency, gravity (seriousness) of reported incidents, and the vulnerability of the victim.

The gap in this approach is that it largely treats incidents as isolated, when they should be looked at as a whole. Research has also found it is used inconsistently between forces.

Most police perpetrator lists contain hundreds or even thousands of people, making them difficult tools to use. They also do not seem to be able to distinguish who the most serious offenders are, with men with very similar profiles near the top, middle and bottom of the lists.

We propose an alternative method, which would assess the whole of a perpetrator’s record of incidents. This would allow police to identify not only the most dangerous perpetrators, but also opportunities to better address their offending earlier on. This might be with diversion to programmes designed to support better choices and rehabilitation, or arrest and incarceration to prevent them harming other people.

By joining together incidents recorded by police for individual perpetrators, we constructed detailed case studies using police officer’s notes. Here is a summary of two people who appear in one force’s perpetrator list.

1. Male born mid 1980s, involved in 340 incidents over 20 years

His offending begins with an indecent assault on a young teenage girl when he is 19. He is increasingly involved in drug-related offending in his 20s. He is later sentenced to six years in jail for arson endangering life. Released on conditional licence, he is re-convicted of the harassment of his ex-partner and recalled to prison.

Release is followed by further offences until the mid-2010s when he is imprisoned again. When released, his offending is erratic (low-level public order, violence, threats, drug-related offending).

Throughout his 30s, he frequently victimises partners and ex-partners. He has no settled address and is homeless at various points of his life. He is still subject to frequent mental health episodes.

2. Male born early 1980s, involved in 396 incidents over 25 years

In his teens he was involved in low-level thefts, criminal damage and breaches of an antisocial behaviour order. He was also suspected of selling drugs to schoolchildren, and imprisoned, aged 18, for drug-related violence.

In his 20s he “associates with” children and is found with a missing vulnerable schoolgirl hiding in his house. He continues to commit offences of criminal damage, drug dealing, and stealing vehicles. Another missing teenage girl is found to be living with him.

In his early 20s he very violently assaults and harasses much younger partners. He continues to commit public order offences and to threaten, harass, and assault current and ex-partners, kicking his pregnant partner in the stomach.

In the early 2020s, police attend his ex-partner’s house following abandoned 999 calls – they find him with his hand over her mouth to stop her calling out to the police. He continues to be violent to ex-partners and his involvement in drug-related offending deepens. He is currently in prison for a violent offence.

Who is the danger?

Both men pose a real and severe threat of violence to women and girls as well as the public. But the RFGV algorithm places the first man more than a thousand places higher than the second. Clearly treating the offences they commit in isolation is not sufficient to distinguish which man poses the greatest risk.

A life-course approach, which takes into account the type and pattern of offending as it develops over time, is less susceptible to fluctuations which move an offender rapidly up or down the priority lists. Therefore, it more reliably reflects who poses the greatest risk.

A better ranking system is clearly required. The RFGV algorithm provides a “score”, but a more sophisticated system would also evaluate the direction of offending of individuals – is it escalating, more frequent, more serious?

A life-course approach could be used separately or together with RFGV to allow police analysts to identify the most serious perpetrators. It may also be possible to use artificial intelligence to identify trends in offending and escalation of risk through analysis of thousands of police incident reports in real time.

The system could then identify opportunities for early intervention which have been shown to be effective in reducing re-offending against current and future victims. It could also automatically trigger warnings to neighbourhood officers, specialist domestic abuse-trained officers, mental health services and so on.

We won’t really know the full capability until new systems are tried, and evaluated. This also means including the voices of survivors and focusing on the lives of persistent perpetrators – often substance use, homelessness, estrangement, imprisonment and mental health problems are at play. The possibilities of learning from artificial intelligence or other technology should not be privileged over the very sources of the data such intelligence relies upon: victims’ experiences.

David Gadd currenty receives research grant funding, via the University of Manchester from the ESRC, NIHR, and Greater Manchester Combined Authority.

Barry Godfrey does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.